[R-sig-Geo] Re: Using adehabitat and splancs to achieve point-in-polygon analysis on raster?

[Sander Oom]

Hi Clement,

Hope you had a good time in the field! Is it spring in France?

Thanks for the very helpful reply! I will try it right now! I need to
present at a conference in early April and would like to get these
results before that!

Just three quick questions before you go to enjoy the Easter weekend:

- how do I get the values out of 'jj' (## jj contains the values of
elevation inside the home range of the animals) to calculate mean,
variance, etc, across all the points in 'jj'? I assume 'jj' is of class
'kasc'. In my case it might be 'asc' as I will have only a single home
range at the time (one animal).

- can I get the points out of the 'kasc' object in a way that I can
still run spatial statistics on them? I would like to calculate the
nugget and range for the NDVI values within the home range.

- do you know how to convert a polygon shapefile into the same class as
produced by the command kernelUD? Alternatively I could import a raster
version of the polygon shape file, convert it to the correct format, and
repeat the spatial join method you are suggesting for the home ranges!

Thanks again,

Sander.


Clément Calenge wrote:
> Dear Sander,
> 
> Sorry for my late answer, I was in the field.
> 
>> I send this email to you directly, because it concerns the link 
>> between Adehabitat and Splancs.
>>
>> I have two sources of polygons:
>> - polygon shape files with areas of interest;
>> - polygons calculated using the home range methods in Adehabitat.
>> For the moment, each polygon source will only contain a single polygon.
>> I have one source of points: a grid representing NDVI values. I read 
>> the grid using the import.asc function in Adehabitat.
>> I would like to use both polygon sources to select the points in the 
>> NDVI raster which are inside the relevant polygons. Then I want to run 
>> several statistics (mean, standard deviation, etc) on the NDVI values 
>> of these points.
> 
> The solution depends on what you call "polygon". The function mcp.rast() 
> is intended to "rasterize"
> a polygon on a raster grid of class "asc" or "kasc". The polygon is here 
> stored as a data frame with
> two columns (the X and Y coordinates of the polygon). The function 
> hr.rast() can be used if you have
> an object of class "area", i.e. a data.frame with three columns (the X 
> and Y coordinates of the polygon,
> and a third column defining the ID of the polygon).
> 
> Now, it seems that  you work with more complex type of polygons, e.g. 
> kernel estimations of home ranges
> (where each animal has an home range, and each home range can be made of 
> several polygons). In such cases,
> the "Spatial join" operation can be more complex. I detail below an 
> example of "spatial join" with a kernel
> estimation of wild boar home ranges. Note, that in this case, you do not 
> have to compute the contour of
> the estimation. Just copy and paste the lines below on R:
> 
> ## loads and display the data
> library(adehabitat)
> data(puechabon)
> 
> ## keep only the elevation
> el<-getkasc(puechabon$kasc,1)
> image(el)
> 
> ## estimate a kernel home-range
> kud<-kernelUD(puechabon$locs[,c("X","Y")], puechabon$locs$Name, grid=el)
> ## kud contains a density
> ## note that the grid containing the values of the elevation is used for 
> the estimation
> 
> vud<-getvolumeUD(kud)
> image(vud)
> ## vud contains the limits of the home ranges estimated at different levels
> 
> toto<-lapply(vud, function(x) x$UD)
> ## toto is the list of maps of Utilization Distribution
> tutu<-lapply(toto, function(x) {x[x<95]<-1; x[x>95]<-NA; return(x)})
> ## tutu is a list of raster maps containing 1 if the pixel is inside
> ## the 95% home range and NA otherwise
> 
> foo<-as.kasc(tutu)
> image(foo)
> ## foo is a kasc object
> 
> jj=lapply(foo, function(x) x*el)
> image(jj)
> ## jj contains the values of elevation inside the home range of the animals
> 
> The last method that is available for home-range estimation of animals 
> home ranges
> is the Nearest Neighbor Convex hull method of Getz and Wilmers (2004). 
> The function
> NNCH.rast() can be used to "rasterize" the home range estimation in the 
> same way, and then
> you can multiply the resulting maps with your NDVI in the same way.
> Hope this helps,
> 
> 
> Clément.
> 

-- 
--------------------------------------------
Dr. Sander P. Oom
Animal, Plant and Environmental Sciences,
University of the Witwatersrand
Private Bag 3, Wits 2050, South Africa
Tel (work)      +27 (0)11 717 64 04
Tel (home)      +27 (0)18 297 44 51
Fax             +27 (0)18 299 24 64
Email   sander at oomvanlieshout.net
Web     www.oomvanlieshout.net/sander